RU2672121C2 - Method and system for controlling tuning factor due to sensor replacement for closed-loop controller in artificial pancreas - Google Patents

Abstract

FIELD: medicine.

SUBSTANCE: group of inventions relates to medical equipment. Method of controlling a system for treating diabetes is carried out using a controller with a predictive model. System comprises an infusion pump with a microcontroller for controlling the pump and obtaining data from a glucose sensor. Method includes measuring glucose concentration, determining a change in system accuracy, which includes determining a decrease in system accuracy in response to the fact that the glucose sensor has been replaced by a new glucose sensor within a predetermined time interval, and determining the increase in system accuracy in response to the calibration of the glucose sensor, setting a tuning factor for the controller with a predictive model in the microcontroller for an overstated tuning factor in response to determining a decrease in system accuracy or an aggressive tuning factor in response to determining an increase in system accuracy. When calculating the amount of insulin, a tuning factor is used to calculate the amount of insulin to be administered to the patient throughout one or more discrete time intervals. Amount of insulin determined at the calculation stage is administered by activating the pump using a microcontroller. System for treating diabetes and a method of controlling a system for treating diabetes using a controller with a predictive model are disclosed.

EFFECT: technical result consists in reducing inaccuracies in the continuous measurement of glucose.

15 cl, 3 dwg

Description

BACKGROUND

Diabetes mellitus is a chronic metabolic disorder caused by the inability of the body to produce enough insulin hormone, which leads to a decrease in the body's ability to absorb glucose. Such inability leads to hyperglycemia, i.e. excess glucose in plasma. Persistent hyperglycemia and / or hypoinsulinemia are associated with various serious symptoms and long-term, life-threatening complications, such as dehydration, ketoacidosis, diabetic coma, cardiovascular disease, chronic renal failure, retinal damage and nerve damage with the risk of limb amputation. Since it is not yet possible to restore the production of endogenous insulin, continuous therapy is necessary to ensure continuous glycemic control so that the concentration of glucose in the blood always remains within normal limits. This glycemic control is achieved by regularly introducing insulin from the outside into the patient's body in order to reduce high glucose.

Exogenous biological products, such as insulin, are usually prescribed as a mixture of fast-acting drugs and drugs with an intermediate duration of action, administered by repeated daily subcutaneous injections with a syringe. It was found that the degree of glycemia control achievable in this way is lower than optimal, since such drug delivery differs from the physiological production of the hormone when the hormone enters the bloodstream at a slower rate and over a longer time. It is possible to improve glycemic control with the help of so-called intensive hormone therapy, based on repeated daily injections, which include one or two injections of a long-acting hormone per day to create a basal concentration of the hormone and additional injections of the fast-acting hormone before each meal in an amount proportional to Serving Size Although insulin pen syringes have at least partially replaced traditional syringes, frequent injections are very uncomfortable for patients, especially those who are not able to inject themselves properly.

The development of a drug delivery device has significantly improved the treatment of diabetes, eliminating the need for multiple daily injections with syringes or syringe pens. The device for drug delivery allows you to enter the drug in a manner closer to the natural physiological processes, and to improve glycemic control in a patient can be regulated in accordance with standard or individual protocols.

In addition, drug delivery devices allow the administration of drugs directly into the intraperitoneal space or intravenously. Drug delivery devices can be designed as implantable devices for placement under the skin or as external devices with an infusion system for subcutaneous infusion by means of a transdermal catheter, cannula or transdermal drug transport, for example, through a patch. External drug dispensers are placed on top of clothing, hidden under or inside clothing, or placed on the surface of the body and are usually controlled using the built-in user interface or a separate remote control device.

To achieve acceptable glycemic control, periodic measurement of the concentration of glucose in the blood or tissue fluid is necessary. For example, delivering the appropriate amount of insulin using a drug delivery device requires the patient to regularly determine the concentration of glucose in his blood and manually enter this value using the user interface for external pumps, which then calculates the required change in the default insulin administration protocol or the current protocol, those. dosage and time of administration, and interacts with a device for drug delivery in order to adjust its operation accordingly. Occasional measuring devices, such as hand-held electronic analyzers using enzyme-linked blood test strips and calculating glucose concentration based on an enzymatic reaction, are typically used to determine blood glucose concentrations in the blood.

Over the past 20 years, continuous glucose control has also been used in combination with drug delivery devices, which allows you to control the infusion of patients with insulin diabetes using a feedback mechanism. To provide the ability to control insulin infusion, proportional-integral-differential (PID) controllers are used with a mathematical model of the interaction between glucose and insulin in humans. PID controllers can be configured based on simple metabolic model rules. However, if you configure or tune PID controllers to aggressively adjust the patient’s blood glucose concentration, an overdose may occur that exceeds the set level, often followed by fluctuations that are highly undesirable in the context of regulating blood glucose concentration. Alternative controllers have been investigated. It is established that a controller with a predictive model (KPM), used in the petrochemical industry and characterized by significant time delays and system responses, is most suitable for adjusting the complex relationship between insulin, glucagon and glucose in the blood. It is shown that KPM is more stable than PID, since KPM takes into account the effects of changes in the near future and restrictions when determining the KPM output, while PID usually takes into account only previous outputs when determining future changes. Restrictions can be used in the KPM, for example, so that the KPM prevents the system from getting out of control when the limit has already been reached. Another advantage of KPM is that the model in KPM can theoretically compensate for dynamic changes in the system in some cases, while when adjusting by the feedback principle, such as PID control, such dynamic compensation is not possible.

Thus, KPM can be considered as a combination of regulation on the basis of feedback and advanced regulation. However, KPM usually requires a metabolic model that imitates as closely as possible the relationship between insulin and glucose in the biological system. Thus, due to individual biological differences, the process of developing and refining KPMs is still ongoing. As an overview of KPM, relevant to the details of KPM, the following documents show and describe KPM options and mathematical models representing the complex interactions between glucose and insulin:

U.S. Patent No. 7,060,059;

US patent application No. 2011/0313680 and 2011/0257627;

International publication WO 2012/051344;

Percival et al., “Closed-Loop Control and Advisory Mode Evaluation of an Artificial Pancreatic β Cell: Use of Proportional-Integral-Derivative Equivalent Model-Based Controllers” Journal of Diabetes Science and Technology, Vol. 2, Issue 4, July 2008.

Paola Soru et al .., “ MPC Based Artificial Pancreas; Strategies for Individualization and Meal Compensation ” Annual Reviews in Control 36, p.118-128 (2012),

Cobelli et al., “Artificial Pancreas: Past, Present, Future” Diabetes Vol. 60, Nov. 2011;

Magni et al., “ Run-to-Run Tuning of Model Predictive Control for Type 1 Diabetes Subjects: In Silico Trial” Journal of Diabetes Science and Technology, Vol. 3, Issue 5, September 2009;

Lee et al., “A Closed-Loop Artificial Pancreas Using Model Predictive Control and a Sliding Meal Size Estimator” Journal of Diabetes Science and Technology, Vol. 3, Issue 5, September 2009;

Lee et al., “A Closed-Loop Artificial Pancreas based on MPC: Human Friendly Identification and Automatic Meal Disturbance Rejection” Proceedings of the 17th World Congress, The International Federation of Automatic Control, Seoul Korea July 6-11, 2008;

Magni et al., “Model Predictive Control of Type 1 Diabetes: An in Silico Trial” Journal of Diabetes Science and Technology, Vol. 1, Issue 6, November 2007;

Wang et al. , "Automatic Bolus and Adaptive Basal Algorithm for the Artificial Pancreatic β-Cell" Diabetes Technology and Therapeutics, Vol. 12, No. 11, 2010; and

Percival et al ., “Closed-Loop Control of an Artificial Pancreatic β-Cell Using Multi-Parametric Model Predictive Control,” Diabetes Research 2008.

All articles or documents cited in this application are incorporated herein in their entirety by reference.

SUMMARY OF THE INVENTION

The authors of the application developed a technique that allows you to configure the predicted control model so that the inaccuracies inherent in the glucose sensor for continuous measurement can be qualitatively compensated in the controller with the predictive model that is used in the microcontroller of this system. In particular, the method is used to control an infusion pump using a microcontroller designed to control the pump and obtain data from at least one glucose sensor. The method can be implemented by: measuring a patient’s glucose concentration using a glucose sensor to provide at least one glucose measurement for each time interval from a series of indexed discrete time intervals (k); determining whether the glucose sensor has been replaced by a new glucose sensor within a predetermined time interval; if the answer at the stage of determining “yes”, setting the tuning factor (R) for the controller with the predictive model in the microcontroller to an overestimated value, otherwise, if the answer at the stage of determining “no”, maintaining the current value of the tuning coefficient (R) for the controller; calculating the amount of insulin for delivery using a microcontroller based on a controller with a predictive model that uses a tuning factor (R) to calculate the amount of insulin for administration to the patient over one or more discrete time intervals; and introducing the amount of insulin determined in the calculation step by activating the pump using a microcontroller.

In yet another aspect, a diabetes control system is provided that comprises a glucometer for occasional measurement, a glucometer for continuous measurement, and an infusion pump connected to the controller. The glucometer for occasional measurement is configured to determine the glucose concentration in the patient’s blood at discrete and non-uniform time intervals and to present a similar episodic concentration of glucose in the blood in the form of a calibration. The glucose meter for continuous measurement is configured to continuously measure the patient’s glucose level at discrete and usually even time intervals and represent glucose concentrations for each interval as glucose measurement data. The insulin infusion pump is configured to deliver insulin. The microcontroller is connected to a pump, a glucometer and a continuous meter. In particular, the controller sets the tuning factor (R) to an overestimated value for a given time interval after the glucometer for continuous measurement is replaced by a new glucometer for continuous measurement, so that the controller determines the speed of insulin delivery for each time interval in the indexed time interval (k ) from the predicted control model based on the overestimated value of the tuning factor and gives the pump a command to deliver at a certain speed of insu delivery ins.

In any of the aspects described above, the following elements may also be used in combination with each of the aspects. For example, at least one glucose sensor may include both a glucose sensor for continuous measurement and a glucose meter for occasional measurement; an overestimated tuning factor (R) may have a value of approximately 500, and a predetermined time interval may be approximately 24 hours from the moment the glucose sensor is replaced with a new glucose sensor; an aggressive tuning factor (R) may have a value of approximately 10; an aggressive tuning factor (R) may have a value of approximately 1000; a controller with a predictive model has an error weighting factor (Q) that is related to a tuning factor (R), where:

Where R may include a tuning factor;

Q may include an error weighting factor.

These and other embodiments, features, and advantages will become apparent to those skilled in the art after studying the more detailed description below of various embodiments of the present invention in conjunction with the accompanying drawings, which are briefly described at the beginning of the application.

BRIEF DESCRIPTION OF THE FIGURES

The accompanying figures, which are included in this document and constitute an integral part of the present description, illustrate the currently preferred embodiments of the invention and, together with the above general description and the detailed description below, serve to explain the elements of the invention (moreover, similar elements are presented under similar numbers )

In FIG. 1 shows a system in which the pump controller or glucose sensor (s) is separated from the infusion pump and the glucose sensor (s) and in which the controller can be connected to the network for real-time monitoring.

In FIG. 2A shows an exemplary embodiment of a system for treating diabetes in a schematic form.

In FIG. 2B shows a curve of glucose concentration values for a time interval of 48 hours (or in the context of k = 12 for each hour of a NKG signal), where other events, such as loss of NKG data, sensor replacement or calibration measurements, are superimposed on a curve of glucose values.

In FIG. 2C shows the tuning factor for a 48-hour time interval (or in the context of k = 12 time interval or index for each hour of the TAG signal), the setting factor R depending on the loss of TCH data, sensor replacement, and calibration measurement.

In FIG. 3 shows the logic used in the controller of FIG. 1 or 2A.

MODES FOR CARRYING OUT THE INVENTION

The detailed description below should be interpreted with reference to the figures in which similar elements in different figures are numbered identically. The figures, not necessarily drawn to scale, show selected embodiments and are not intended to limit the scope of the present invention. In a detailed description, the principles of the invention are shown by way of examples, which are not restrictive. The present description will undoubtedly allow specialists in this field to realize and apply the invention, and it presents several embodiments, adaptations, variations, alternatives and applications of the invention, including those that are currently considered the best options for implementing the invention.

As used herein, the term “approximately” with respect to any numerical values or ranges indicates a suitable dimensional tolerance that allows a part or a combination of components to perform the function provided for them in this document. In addition, in this document, the terms “patient”, “operator”, “user” and “subject” refer to any human subject or animal subject and do not imply a restriction on the use of systems or methods only in humans, although the patient uses the subject matter -man is a preferred embodiment. In addition, the term “user” includes not only the patient using the infusion pump to administer the drug, but also those responsible for it (for example, parents or guardians, nursing staff, or home care workers). The term “drug” can include a hormone, biologically active materials, pharmaceuticals, or other chemicals that cause a biological reaction (eg, glycemic reaction) in the body of a user or patient.

In FIG. 1 shows a drug delivery system 100 in accordance with an embodiment using the principles of the present invention. The drug delivery system 100 includes a drug delivery device 102 and a remote controller 104. The drug delivery device 102 is connected to the infusion set 106 via a flexible tube 108.

The drug supply device 102 is configured to transmit data to and receive data from the remote controller 104, for example, by radio frequency communication 112. The drug delivery device 102 can also function as a standalone device with its own built-in microcontroller. In one embodiment, the drug delivery device 102 is an insulin infusion device, and the remote controller 104 is a hand held portable controller. In such an embodiment, the data transmitted from the drug delivery device 102 to the remote controller 104 may include information such as, for example, insulin delivery data, information on blood glucose concentration, basal concentration, bolus administration, insulin to carbohydrate ratio or insulin sensitivity coefficient, as a few examples. The configuration of the microcontroller 104 includes a KPM controller 10 programmed to continuously obtain glucose measurement results from a sensor for continuously measuring blood glucose concentration 112. Data transmitted from the remote microcontroller 104 to the insulin delivery device 102 may include glucose measurement results and a food intake database so that the drug delivery device 102 can calculate the amount of insulin to be delivered by the delivery device 102. Alternatively, the remote microcontroller 104 may perform basal dose or bolus dose calculations and send the results of such calculations to the drug delivery device. In an alternate embodiment, the occasional meter glucometer 114 may be used alone or in combination with a NKG sensor 112, providing data to the microcontroller 104 and / or the drug delivery device 102. Alternatively, the remote microcontroller 104 may be combined with meter 114 to form either (a) an integrated one-piece device; or (b) two separate devices that can be connected to each other to form an integrated device. Each of the devices 102, 104, and 114 has a corresponding microcontroller (not shown for brevity) programmed to perform various functions.

The drug delivery device 102 may also be configured for bidirectional wireless communication with a remote health monitoring station 116 via, for example, a wireless communication network 118. Remote controller 104 and a remote monitoring station 116 can be configured for bidirectional wired communication through, for example, a communication network based on the telephone line. Remote monitoring station 116 may be used, for example, to download updated software to a drug supply device 102 and to process information from a drug supply device 102. Examples of remote monitoring stations 116 may include, but are not limited to, a personal or network computer 126, a server 128 with memory for storage, a personal digital assistant, another mobile phone, a base station for monitoring in a hospital, or a special remote station for clinical monitoring.

The drug delivery device 102 includes components for processing an electronic signal, including a central processor and memory elements for storing control programs and operational data, a radio frequency module 116 for sending and receiving radio communication signals (i.e., messages) to a remote controller 104 or from it, a display for displaying operating information for the user, a plurality of navigation buttons allowing the user to enter information, a battery for supplying system power, presentations (visual, audible or tactile) for user feedback, a vibrating device for user feedback, a drug delivery mechanism (e.g., a pump and a drive mechanism) for displacing insulin from an insulin reservoir (e.g., an insulin cartridge) through a side opening , to which the 108/106 infusion system is connected, to the user's body. An example of a drug delivery device 102 (or pump 16) may be provided in the form of a modified Animas Vibe insulin pump manufactured by Animas Corporation, Wayne, pc. Pennsylvania, USA

Glucose levels or concentrations can be determined using the 112 NCT sensor. The sensor 112 for continuous glucose monitoring uses amperometric electrochemical measurement technology to determine glucose using three electrodes that are functionally connected to the sensor electronics and coated with a sensitive membrane and a biointerface membrane that are connected by a clamp.

The upper ends of the electrodes are in contact with the electrolyte phase (not shown), which is a free-floating liquid phase located between the sensitive membrane and the electrodes. The sensitive membrane may include an enzyme (e.g., glucose oxidase) that covers the electrolyte phase. In this example of a sensor, a counter electrode is proposed for balancing the current generated by a substance when measured on a working electrode. In the case of a glucose oxidase-based glucose sensor, the substance measured at the working electrode is H ₂ O ₂ . The current received at the working electrode (and flowing through the circuit to the counter electrode) is proportional to the diffusion flux of H ₂ O ₂ produced by the electromechanical conversion of glucose into an enzymatic by-product. Accordingly, an untreated signal can be obtained that represents the concentration of glucose in the patient’s body and therefore can be used to estimate the meaningful glucose value. Details of the sensor and related components are shown and described in US Pat. No. 7,276,029, which is incorporated herein by reference in its entirety. In one embodiment, a continuous glucose sensor from a Dexcom Seven® system (manufactured by Dexcom Inc.) may be used with the embodiments described herein.

The following components may also be used as an artificial pancreatic diabetes treatment system in one embodiment: Animas Corporation's OneTouch Ping® glucose concentration control system comprising at least an infusion pump and an occasional glucose sensor; and DexCom® G4 Platinum® CGM manufactured by DexCom Corporation with an interface for connecting these components and programming using MATLAB® language and additional hardware for connecting components to each other; as well as KPM control algorithms that automatically adjust the speed of insulin delivery based on the patient’s glucose concentration, past glucose measurement results, predicted future trends in glucose concentration and specific patient information.

In FIG. 2A shows a circuit 200 of the system 100 of FIG. 1, programmed using a solution developed by the applicants to prevent a feedback control system from decreasing below the desired effect. In particular, in FIG. 2A shows the KPM programmed as a logical control module 10 used in the controller 104. The KPM logical control module 10 obtains the required glucose concentration or glucose concentration range 12 (along with any modification from the update filter 28) so as to be able to support the output signal (t ie the glucose concentration) of the patient in the required range of glucose concentration.

As shown in FIG. 2A, the first output 14 of the CPM-activated logic control device 10 may be a control signal for the insulin pump 16 to deliver the desired amount of insulin 18 to the patient 20 at predetermined time intervals that can be indexed every 5 minutes using the time interval index k. A second output, in the form of a predicted glucose concentration value 15, can be used in control compound B. Glucose sensor 22 (or 112 in FIG. 1) measures glucose concentration in patient 20 to provide signals 24, reflecting the actual or measured glucose concentration, to control compound B , which takes into account the difference between the result of measuring glucose concentration 24 and KPM forecasts regarding this measured glucose concentration. Based on this difference, information is supplied to the update filter 26 of the model state variables. Difference 26 serves as an assessment parameter (also update filter 28) and allows you to evaluate state variables in the model that cannot be measured directly. Preferably, the update filter 28 is a recursive filter in the form of a Kalman filter with parameter tuning for the model. The output signal of the update filter 28, or recursive filter, is transmitted to the control connection A, the output of which is used by the CPM in the logic control device 10 to further refine the control signal 14 to the pump 16 (or 102 in Fig. 1). A tuning factor of 34 is used with the KPM 10 controller to “tune” the controller during insulin delivery. To achieve this, applicants have developed the use of a calibration index module 30 and a data skip module 32 to adjust the tuning factor. Calibration index module 30 is configured to track the number of glucose calibration calibrations that is typically performed by a glucometer for occasional measurement, such as, for example, a test strip for determining blood glucose and a measurement system. The data skip module 32 is configured to track the number of missed measurements or data received from the meter 22 for continuous glucose measurement.

Here it is necessary to provide a brief overview of the aforementioned KPM, which uses the logical control device 10. The CPM logic is formulated in such a way that it provides control of the patient’s glucose concentration within the safe glucose value zone, with the lower limit of blood glucose concentration in this zone being 80–100 mg / dl, and the upper limit of blood glucose concentration being 140–180 mg / dl; hereinafter, this algorithm will be called the KPM zone. In general, the control of the desired zone is applied to controlled systems in which there is no specific set value, while the purpose of the controllers is to maintain a controlled variable (KP), for example, glucose value, within the set zone. Monitoring to maintain in a specific area (i.e., the normoglycemic zone) is very suitable for an artificial pancreas due to the lack of an established glycemic value in vivo. In addition, the advantage of the control method in a certain zone is the ability to limit the response / activity of the pump in such a way that, while maintaining the glucose concentration within the zone, no additional correction is proposed.

In real time, the rule that regulates the rate of delivery of insulin I _D in the KPM zone is calculated by real-time optimization, and the speed of delivery of the next dose of insulin is estimated at each time the sample is received. Optimization at each time point of the sampling is based on the assessment of the metabolic state (plasma glucose concentration, subcutaneous administration of insulin) obtained from a dynamic model stored in module 10.

, где

- это период отбора пробы. Программные ограничения гарантируют, что скорость доставки инсулина остается в пределах от минимального (например, нуль) до максимального значений. Затем вводится первая инфузия инсулина (вне соответствия

шагам). На следующем шаге,

на основании нового результата измерения глюкозы и последней скорости введения инсулина, процесс повторяется.The CPM logical control device 10 includes a detailed model of the dynamics of glucose-insulin in humans with type I diabetes. The model is used to predict future glucose concentrations and to calculate future controller actions to bring the glucose concentration graph to the desired range. KPM in controllers can be programmed for both discrete and continuous systems; the controller is set to discrete time, while the index of discrete time (stage) k refers to the k- ^th sampling period of the sample for continuous time

where

is the sampling period. Software constraints ensure that insulin delivery rate stays between a minimum (e.g., zero) and a maximum value. Then the first insulin infusion is administered (out of line

steps). In the next step,

based on a new glucose measurement result and the last rate of insulin administration, the process repeats.

In this regard, we start with the initial linear model of differences used for the KPM zone:

Where:

k is an index of a discrete time interval having a series of index counters, where k = 1, 2, 3 ...

G ’is a measured glucose concentration

I _M is “insulin in the schedule of administration”, the amount of which has not been measured;

I ' _D is an insulin delivered or controlled variable;

and coefficients a ₁ ~ 2.993; a ₂ ~ (-3.775); a ₃ ~ 2.568; a ₄ ~ (-0.886); a ₅ ~ 0.09776; b ~ (-1.5); c ₁ ~ 1.665; c ₂ ~ (-0.693); d ₁ ~ 0.01476; d ₂ ~ 0.01306.

Using an FDA-approved metabolic model known to those skilled in the art, ur. (1) can be simplified to the following linear model of differences in ur. (2):

(2)

Where:

G ' is a variable deviation (mg / dl) of glucose concentration output ( G ), i.e.

I _D ' is a variable of the deviation (U / h) of the input of the insulin infusion rate ( I _D ), i.e.

Eating is the intake of carbohydrate intake (grams of carbohydrates),

I _M is the rate of subcutaneous insulin infusion in accordance with the scheme of administration (U / h),

Meal _M is the intake of carbohydrate intake in accordance with the scheme of administration (grams of carbohydrates).

Dynamic model in ur (2) refers to the effect of insulin infusion rate ( I _D ) and the intake of carbohydrates consumed ( Meal ) on plasma glucose concentration. This model is a single average model for the general population of subjects. The model and its parameters are fixed.

Transfer functions of second-order inputs described in parts (b) and (c) in ur. (2), are used to generate artificial information about memory entries when using the KPM zone scheme to prevent insulin overdose and, therefore, hypoglycemia. In order to avoid the delivery of excessive amounts of insulin in assessing any sequential administration of insulin, it is necessary to take into account the last time insulin is compared with the duration of action of insulin. However, a linear model of distinguishing one state with a relatively low order uses output (glycemia) as the main source of previous inputs (insulin) in memory. Due to the mismatch of the model, noise or a change in the patient's sensitivity to insulin is possible, which can lead to the delivery of insufficient or excessive amounts of insulin. This effect can be reduced by adding two additional states (I _M and Eating _M) for insulin injections included in the scheme and meals with a long memory of insulin.

The KPM zone is used when the specific set value of the controlled variable (KP) has low significance compared to the zone defined by the upper and lower limits. In addition, in the presence of noise and model mismatch, the use of a fixed installation is not practical. The KPM zone was developed through research at the University of California, Santa Barbara and the Sansuma Diabetes Research Institute. Detailed information on the development of the KPM zone technique is presented and described in the following sources: Benyamin Grosman, Ph.D., Eyal Dassau, Ph.D., Howard C. Zisser, MD, Lois Jovanovič, MD, and Francis J. Doyle III, Ph.D. “ Zone Model Predictive Control: A Strategy to Minimize Hyper and Hypoglycemic Events ” Journal of Diabetes Science and Technology, Vol. 4, Issue 4, July 2010, and the publication of US patent application No. 2011/0208156, owned by Doyle et al., Entitled " Systems, Devices, and Methods to Deliver Biological Factors or Drugs to a Subject ", which was published on August 25 2011 and is incorporated herein by reference with a copy in the appendix. Additional information about the KPM zone is shown and described in the publication of US patent application No. 20110208156, incorporated herein by reference with a copy in the appendix. A similar development of the KPM zone was presented by Maciejowski JM., “ Predictive Control with Constraints” Harlow, UK: Prentice-Hall, Pearson Education Limited, 2002. The KPM zone is introduced by setting fixed upper and lower boundaries as flexible limiting values. so that the values during optimization switch between zero and some final values when the predicted CPs are inside or outside the desired zone, respectively. Predicted differences are usually defined as the difference between a CP beyond the desired zone and the closest boundary. The KPM zone is usually divided into three different zones. The allowed range is the control objective and is determined by the upper and lower limits. The upper zone represents undesirable high predicted glycemia values. The lower zone represents undesirable low predicted glycemic values, which represents the hypoglycemia zone or protective area before hypoglycemia, which is considered the lower warning zone. The KPM zone optimizes predicted glycemia by manipulating actions aimed at controlling insulin in the near future so that the concentration remains in the allowed zone limited by the indicated values.

The essence of the KPM zone lies in the target functional wording, which contains the wording of the zone. The KPM zone, like any other form of KPM, predicts future results using a detailed model using past I / O data and future input signals that need to be optimized. However, instead of leading to a specific fixed set value, optimization attempts are aimed at maintaining or shifting the predicted results into the zone defined by the upper and lower boundaries. Using the linear difference model, the dynamics of glycemia is predicted and optimization is carried out, which reduces future glycemic fluctuations outside the zone defined by the restrictions and values specified in its cost function.

The objective function J of the KPM zone used in the present work is designated as follows:

or applicable to our task:

Where

J is the objective function;

Q is the weighting coefficient for the predicted glucose value;

R is the tuning factor for estimated future inputs within the objective function;

f is a predictive function (in level (2)); and

vector I _D contains a set of insulin doses for infusion, offered in the near future. It is a "controlled variable" because it is adjusted in accordance with the search for the minimum value of J.

The G- ^zone is a variable that quantitatively expresses the deviation of the G values predicted in the model of the future NCG from the limits of a specific glycemic zone and determined by the following comparisons:

where the glycemic zone is limited by the upper limit of G _ZH and the lower limit of G _ZL .

Thus, if all the predicted glucose values are within the zone, then each element of the G ^zone is equal to 0, as a result of which the function J is minimized under the basal mode I _D for this time of day, i.e. for patients with current basal insulin infusion rate, the default algorithm is used. On the other hand, if any of the predicted glucose values is outside the zone, then the G- ^{zone is} > 0, which affects the target function. In this case, the doses proposed in the near future insulin infusion I _D will differ from basal to prevent any deviation G ^-zone of the zone, which also affects the objective function. A quantitative balance is then found using optimization based on the weighting coefficient R.

To solve problems with optimization of equations (2) - (5), commercially available software is used (for example, the “fmincon.m” MATLAB function). When using this function, the following parameters are used for each optimization:

, например, если M=5, то начальное приближение для каждой оптимизации составляет I _D ’=[0 0 0 0 0]. Это означает, что начальное приближение эквивалентно базальной скорости.o The initial approximation for the rate of insulin delivery I _D ' (0) is the zero vector

for example, if M = 5, then the initial approximation for each optimization is I _D '= [0 0 0 0 0]. This means that the initial approximation is equivalent to basal velocity.

o The maximum number of valid estimates using the function Max_f = 100 * M , where M is the control interval described earlier.

o The maximum number of iterations Max_i = 400, it is fixed.

o Termination of the calculation of the objective function occurs when the value Term_cost = 1e-6, it is fixed.

○ The termination of the calculation of the deviation Term_tol for the controlled variables I _D 'is 1e-6.

The following stringent restrictions apply to controlled variables ( I _D '):

where the basal regimen is the basal delivery rate that has been established by the patient or his / her physician,

expected values range from about 0.6 to about 1.8 U / h.

Although the values of the parameter of the control interval M and the parameter of the estimated interval P significantly affect the performance of the controller and are normally used to configure the controller based on the KPM, they can be configured heuristically based on the knowledge of the system. Setup rules are well known to those skilled in the art. In accordance with these rules, M and P can take values in the following ranges:

In preferred embodiments, we used nominal values of M = 5 and P = 108.

The ratio of the weighting coefficient of the error of the output quantity Q and the weight matrix of the input changes or the tuning factor R can take values in the range:

In preferred embodiments, we can use a nominal value of R / Q = 500.

After turning on and starting the controller 10 in the microcontroller in the device 102 or 104, real-time calculations are performed every 5 minutes in accordance with the sampling time for measurement by the glucose sensor 22. The first element I _{D is} delivered to the patient in the form of a dose of insulin using an insulin pump 16, after after five minutes, a new reading will be available within the framework of the NCT and the process is repeated. It should be noted that future actions of the controller have severe limitations due to the ability of the insulin pump to deliver insulin at maximum speed or the inability to deliver negative insulin values. Other details related to the subject, including condition assessment and other CPMs, are described here: Rachel Gillis et al., “ Glucose Estimation and Prediction through Meal Responses Using Ambulatory Subject Data for Advisory Mode Model Predictive Control ” Journal of Diabetes Science and Technology Vol. 1, Issue 6, Nov. 2007, and Youqing Wang et al., " Closed-Loop Control of Artificial Pancreatic β-Cell in Type 1 Diabetes Mellitus Using Model Predictive Iterative Learning Control " IEEE Transactions on Biomedical Engineering, Vol. 57, No. 2, February 2010, which are fully incorporated into this application by reference.

It is known that a setting or a setting factor (referred to herein as “R”) can have a significant effect on glucose control quality. The parameter, known among other names as the aggressiveness factor, gain, etc., determines the response rate of the algorithm to changes in glucose concentration. With a relatively high R value, the controller slows down to regulate the insulin dose for infusion (relatively basal) in response to changes in glucose concentration; on the other hand, the relatively aggressive value of R causes the controller to respond quickly to changes in glucose concentration. In principle, an aggressive controller may lead to better glucose control if 1) the available glucose measurements are accurate and, moreover, 2) the predictive model of future changes in glucose concentration is accurate. If these conditions are not met, it is safer to use a controller with overestimated values of the coefficient.

A discussion of tuning factor R (referred to here as 34) in FIG. 2A is appropriate at this stage. If it is difficult to determine the accuracy of online glucose measurements or the accuracy of model predictions, it may be difficult to determine whether to use factor 34 or R under these conditions. Nevertheless, under certain circumstances it can be established with a high degree of confidence if continuous glucose monitoring (NCT) has become “less accurate” or “more accurate” from a quantitative point of view.

Applicants consider it reasonable to assume that NKG may become less accurate when analyzing each subsequent sample after calibration, and NKG will suddenly become less accurate when it cannot report the result of glucose measurements. Lost or missed measurements with NKG may be due to uncertain changes in glucose concentration (a feature of the NKG algorithm) or simply because of problems with the transmission of data via radio frequency communication. Similarly, it is reasonable to assume that the NCH will become more accurate after calibrating and recovering messages with glucose values after missing one or more reads.

Then, the authors of the application found that it is possible to use these ideas to automatically configure the controller, i.e. use an overestimated R value for the tuning factor when the NKG becomes less accurate, and a more aggressive value when the NKG becomes more accurate. In a general sense, this is presented below.

Let the aggressiveness factor or tuning factor R be limited by the constant Rmin (for the most aggressive controller operation mode) and Rmax (for the least aggressive controller operation mode):

Rmin <= R <= Rmax ur. (9)

Let there be two nominal increments of a strictly positive value: r ₁ , a relatively small increment associated with the accuracy of the NCT depending on the time of calibration, and r ₂ , a larger increment associated with the accuracy of the NCT depending on the presence of missing samples.

The tuning factor used in calculating the KPM for the current measurement moment, indexed by the index k, is R (k), where R (k) is automatically updated for each sample based on the nominal value of R, R _HOM , which usually should be different for each user :

R (k) = R _NOM + r ₁ * KAL (k) + r ₂ * SKIP (k), ur. (10)

and R _min <= R (k) <= R _max . (eleven)

KAL (k) = k - k _cal - 6 for k - k _cal > = 6 level (12) KAL (k) = k - k _{cal 2 -} 6 for k - k _cal <6 level (13)

where: R _HOM < R _MAX < 100 * R _HOM

R _HOM / 100 < R _MIN < R _HOM

(R _MAX - R _MIN ) / 500 < r ₁ < (R _MAX - R _MIN ) / 50

(R _MAX - R _MIN ) / 50 < r ₂ < (R _MAX - R _MIN ) / 5

KAL (k) is a calibration index.

It should be noted that R _HOM can take values from 0 to about 1000 (ie 0≤R _HOM ≤1000); although it should be noted that this value and range depend on the specific objective function in the particular model used, and one skilled in the art should be able to configure the appropriate R _HOM range depending on the model used. In exemplary embodiments, r ₁ can take any value from about 1 to 50, and r ₂ can take any value from about 10 to 1000, and preferably about 500. Although the tuning factor is presented as a range of certain numerical values for embodiments described in this document, we note that each system must use different configurations of the tuning factor in order to control the desired response of the controller (and pump) during insulin delivery. Therefore, our intention is to not be tied to any particular numerical value, and thus, as described herein, the tuning factor can be any value from the first value or interval at which the pump will output very fast ( or extremely aggressive) response when delivering insulin, to the second value or interval at which the pump will give a very slow (or overestimated) response when delivering insulin.

The index of the current sample is denoted as “k”, where k _cal is the index of the sample used in the last NCT calibration using a suitable glucose analyzer or glucometer 114, and k _cal2 is the index of the sample used in the penultimate calibration using the 114 glucose control. Thus, after 30 minutes (for example, 6 samples) after calibration with the control glucose meter 114, the module 32 of the calibration index, or KAL (k), is zero. After that, CAL increases by one after each sample before 30 minutes after the next calibration. Such adjustment of the tuning coefficient R of module 34 for the case k (i.e., R (k)) leads to the controller working with slightly more overestimated values (i.e., large R values in equation (4)) for each subsequent measurement of the NKG by the sensor 112 for each indexed time interval k until the next calibration. The presence of a 30-minute buffer period immediately after calibration, during which the logic control device 10 will operate taking into account relatively high values, leads to the possible occurrence of significant differences in the values of the ICG immediately after calibrations.

A detailed description of the setting of the tuning factor for certain events (for example, signal loss or calibration) can be found in US patent application No. 13 / 708,032, filed December 7, 2012, the contents of which are fully incorporated into this application by reference.

In addition to setting the tuning factor for signal loss or calibration of the NKG, we also developed a strategy that is applied when the NKG sensor is inaccurate each time such a sensor is initially placed (or replaced by another sensor) on the user's skin.

Recall that the system of FIG. 2A is proposed for treating diabetes in a patient. The following components are used in this system: a glucometer 29 for occasional measurement, a sensor 22 for continuous glucose measurement, a pump 16 and a controller 10. The glucometer 29 measures the glucose level in the patient’s blood at discrete non-uniform time intervals (for example, every 4 hours) and represents the glucose concentration in blood obtained by occasional measurement, as a calibration index 30 for each time interval in the indexed time interval (k, where the time interval between k and k + 1 is about 5 minutes). The glucose meter for continuous measurement constantly measures the glucose concentration in the patient’s blood at discrete and usually even time intervals (for example, approximately every 30 seconds or every minute) and presents the glucose concentration for each interval as glucose measurement data, in which any measurement is missing glucose at any time interval is stored as a pass index 32. The infusion pump for insulin is controlled by controller 10 to deliver insulin to patient 20. In this document, the value The term “insulin” is not limited to peptide hormone derived from animal tissue, or genetically engineered insulin, biosynthetic recombinant human insulin, or its analogs, but may include other drugs or biological products designed to mimic insulin or other materials, having insulin-like effects on the human body. The controller 10 is programmed to run the appropriate KPM program for controlling the pump and communicating with the meter for occasional measurement and the meter for continuous measurement. In this aspect, the controller determines a tuning factor 34 based on (a) calibration index 30 obtained from occasional glucose measurements and (b) skip index 32 for KPM 10 so that the controller determines the rate of insulin delivery for each time interval in the indexed time interval (k) according to the control predictive model based on (1) the required glucose concentration 12, (2) the glucose concentration 24 measured by the glucometer 22 at each time interval in the indexed time interval (k), and (3) to tuning factor 34. Nevertheless, we developed the system so that in case of changing or replacing the NKG sensor (moreover, the sensor replacement can be detected by the pump or confirmed by the user with the toggle switch), the system sets the value of the tuning coefficient R to an overestimated value, for example, to 500 for time interval, for example, for 8, 12, 24 hours, as well as for longer or shorter intervals, depending on the inaccuracies inherent in the specific NCT sensor used.

As seen in FIG. 2B and 2C, the telemetry outputs of the system are provided to demonstrate the operation of this system. In FIG. 2B, the output of the NCH is presented for a period of 48 hours, while the output of the tuning factor for the same period of time is shown in FIG. 2C after FIG. 2B. In the case of “a” in FIG. 2B, calibration is carried out, which means that in a qualitative sense, the output signal from the NKG sensor can be considered more accurate (due to calibration using a glucose sensor for occasional measurement or a test strip). Therefore, in case “b” in FIG. 2C KPM (in the microcontroller) can increase the aggressiveness of the tuning factor R (reducing its numerical value to 50), and gradually increasing the aggressiveness of the coefficient R over the next two hours (from 12 to 14 hours). In the case of "c" in FIG. 2B, the microcontroller detects using an indicator or toggle switch 36 (or the user confirms this) that the NKG sensor has been changed to a new NKG sensor. The microcontroller, in accordance with the program in the embodiment of our invention, immediately switches the setting factor to an overestimated value, so that the dose of insulin is higher. And for a given time interval from 14 hours to about 40 hours, this overestimated tuning factor is maintained for a specified time (for example, from 14 hours to about 40 hours), regardless of the missing signals from the new NCT sensor. As used herein, the phrase “overshoot factor” refers to a controller coefficient R for the controller, which should ensure that the rate of delivery of the insulin dose (relatively basal in response to changes in glucose concentration for the intended use) is sufficiently low and within the limits of the particular application of the system, to comply with safety and regulatory requirements for such an application. As a non-limiting example, an overestimated gain can have any value from about 500 to 1000. However, it should be clear that this gain can have any value used to influence the response or “gain” of a particular configuration of a system using a controller with feedback (feedback controller, prediction controller or hybrid controller), provided that specialists in this field will consider this setting factor safe or yshennym.

Due to the description provided herein, applicants have also claimed a method for controlling an infusion pump to control an infusion pump using a controller for controlling the pump and receiving data from at least one glucose sensor. An example of a method can be described with reference to FIG. 3. The method begins by measuring a patient’s glucose concentration in step 302 using a glucose sensor (22 or 29) to provide at least one glucose measurement for each time interval from a series of indexed discrete time intervals (k). At step 304, the system asks if there has been a change in the NKG sensor within the specified time interval X. The determination of the sensor change at step 304 can be done by checking the microcontroller’s sensor change log, the indicator of sensor changes, or by combinations of one or more signals indicating a sensor change (for example, manually activated signals by the user). If the answer is “yes” or “yes,” then the system checks at step 306 whether the countdown timer for the given time X is equal to zero (or the forward countdown timer is equal to the specified time X). If the answer is “not true” (ie, “no”), then the system proceeds to step 308, where the value of the tuning factor is set to an overestimated value. At step 312, the KPM in the microcontroller now determines the appropriate insulin dose I _D (from equations 2-5) within the zone, taking into account the objective function J, which is affected by the tuning factor R. At step 314, the microcontroller activates the pump motor in the pump 102 to deliver the appropriate dose the insulin calculated in step 312. In step 316, the microcontroller returns to its normal operation.

On the other hand, if the answer to the request at step 304 is “not true” (ie, “no”), then the system maintains the current value of the tuning factor set before the sensor changes. Alternatively, the system can still calculate the appropriate tuning factor for the workload, as described above for equations 9-13.

Although the present invention has been described in the context of specific modifications and illustrative figures, persons with average skill in the art will recognize that the invention is not limited to the described modifications or figures. For example, a controller with a closed control loop does not have to be a KPM controller, and perhaps, with appropriate modifications by specialists, a PID controller, a PID controller with internal model control (UVM), and model-algorithmic control (UIA), as discussed in the publication Percival et al., “ Closed-Loop Control and Advisory Mode Evaluation of an Artificial Pancreatic β Cell: Use of Proportional-Integral-Derivative Equivalent Model-Based Controllers” Journal of Diabetes Science and Technology, Vol. 2, Issue 4, July 2008. In addition, it will be obvious to those with average skill in the field that in those cases where the above methods and steps indicate the occurrence of certain events in a specific order, this order for some steps may be modified, and that such changes are in accordance with possible embodiments of the present invention. In addition, if possible, certain steps can be performed simultaneously in a parallel process, as well as performed sequentially, as described above. Thus, to the extent that variations of the present invention are possible that are consistent with the essence of the description or are equivalent to the inventions described in the claims, the present patent is intended to cover all such variations as well.

Claims (39)

1. A method of controlling a system for treating diabetes using a controller with a predictive model having a tuning factor (R), wherein the tuning coefficient (R) is adjustable between an overestimated tuning coefficient and an aggressive tuning coefficient, the system comprising an infusion pump with a microcontroller for controlling the pump and receiving data from at least one glucose sensor, the method comprising: measuring a patient’s glucose concentration with a glucose sensor to provide at least one glucose measurement for each time interval from a series of discrete time intervals; determination of changes in the accuracy of the system, and the definition includes: determining a decrease in system accuracy in response to the fact that the glucose sensor has been replaced by a new glucose sensor within a predetermined time interval, and; determining improved system accuracy in response to glucose sensor calibration; setting a tuning factor (R) for a controller with a predictive model in the microcontroller to an overestimated tuning coefficient in response to determining a decrease in system accuracy or an aggressive tuning coefficient in response to determining an increase in system accuracy; calculating the amount of insulin for delivery using a microcontroller based on a controller with a predictive model that uses a tuning factor (R) to calculate the amount of insulin for administration to the patient over one or more discrete time intervals; and the introduction of the amount of insulin, determined at the stage of calculation, by activating the pump using a microcontroller. 2. The method of claim 1, wherein the overestimated tuning factor (R) is approximately 500, and the predetermined time interval is approximately 24 hours from the moment the glucose sensor is replaced with a new glucose sensor. 3. The method of claim 2, wherein the tuning factor (R) has a value of about 10. 4. The method of claim 1, wherein the tuning factor (R) has a value of about 1000. 5. The method according to claim 1, in which the overestimated tuning factor has a value of approximately 500. 6. The method according to p. 1, in which the tuning factor has any value from the first value, at which the pump will give a faster response when delivering insulin, to the second or overvalue, at which the pump will give a slower response when delivering insulin, This overstated setting has an overstated setting factor. 7. The method according to claim 1, in which the controller with the predictive model has an error weighting factor (Q) that is associated with a tuning factor (R), where:

where R is the tuning factor; Q is the error weighting factor. 8. The method of claim 1, wherein the at least one glucose sensor comprises a glucose sensor for continuous measurement and a glucometer for occasional measurement. 9. A system for treating diabetes, comprising: a glucose meter for occasional measurement to determine the concentration of glucose in the blood of a patient at discrete and non-uniform time intervals and presenting this episodic concentration of glucose in the blood as a calibration; a continuous measurement glucometer for continuously measuring a patient’s glucose level at discrete and uniform time intervals and representing the glucose concentration for each interval as glucose measurement data; insulin infusion pump for insulin delivery; a microcontroller using a controller with a predictive model having a tuning factor (R), wherein the tuning coefficient (R) is adjustable between an overestimated tuning coefficient and an aggressive tuning coefficient, the microcontroller being connected to a pump, meter and a continuous meter, in which the controller sets the coefficient settings (R) for an overestimated setting factor for a given time interval after the meter for continuous measurement has been replaced by a new a glucometer for continuous measurement, or an aggressive tuning factor in response to calibrating a glucose sensor for a given time interval after the glucometer for continuous measurement has been calibrated so that the controller determines the speed of insulin delivery for each time interval during time intervals with an index (k ) from the predicted control model based on the tuning factor (R) and sends a command to the pump to deliver insulin at a given delivery speed. 10. The system of claim 9, wherein the overestimated tuning factor (R) has a value of approximately 500. 11. The system of claim 9, wherein the tuning factor (R) has a value of about 10. 12. The system of claim 9, wherein the tuning factor (R) has a value of about 1000. 13. The system of claim 9, wherein the controller with the predictive model has an error weighting factor (Q) that is associated with a tuning factor (R), where:

where R is the tuning factor; Q is the error weighting factor. 14. The system according to claim 9, in which the tuning factor has any value from the first value at which the pump will give a quicker response when delivering insulin, to the second or overvalue, at which the pump will give a slower response when delivering insulin, This overstated setting has an overstated setting factor. 15. A method of controlling a system for treating diabetes using a predictive model controller having a tuning factor (R), wherein the tuning coefficient (R) is adjustable between an overestimated tuning coefficient and an aggressive tuning coefficient, the system comprising an infusion pump with a microcontroller for controlling the pump and receiving data from at least one glucose sensor, the method comprising: measuring a patient’s glucose concentration with a glucose sensor to provide at least one glucose measurement for each time interval from a series of discrete time intervals; determining a change in system accuracy, the definition including: determining a decrease in system accuracy in response to the fact that at least one glucose sensor has been replaced with a new glucose sensor within a specified time interval, with an increase in the number of samples when analyzing each subsequent sample after calibrating the glucose sensor or if the glucose sensor fails to report the glucose measurement result; determining an increase in system accuracy in response to one or more of a calibration of a glucose sensor or restoring a message by a glucose sensor to glucose measurement results; setting a tuning factor (R) for a controller with a predictive model in the microcontroller to an overestimated tuning coefficient in response to determining a decrease in system accuracy or an aggressive tuning coefficient in response to determining an increase in system accuracy; calculating the amount of insulin for delivery using a microcontroller based on a controller with a predictive model that uses a tuning factor (R) to calculate the amount of insulin for administration to the patient over one or more discrete time intervals; and the introduction of the amount of insulin, determined at the calculation stage, by activating the pump using a microcontroller.

Images